The DRAM price meltdown seems to have hit bottom, at least for the short term. Amazingly, during February and early March, 128Mb SDRAM chip prices plummeted to almost what 64Mb prices were this time last year – just over $4 per chip. Most recently, prices have firmed a bit, and are now approaching $5, though that is still below manufacturing cost for some chipmakers.
Last year, memory manufacturers were able to find some relief by converting to Flash and specialty SRAMs, however now Flash memory prices are plummeting as well, with prices more than 30% below their levels of last quarter. These conditions have caused worries to resurface that some manufacturers may not be able to survive in the DRAM market.
In this environment, anything that allows parts to be produced cheaper is going to gain support, and DDR SDRAM is definitely one of those items. It has been stated by experts in memory design and production that not only can DDR SDRAM be produced on the same product line, it can actually be supported on the same chip with a little extra circuitry. Since the determination of whether a chip is SDR or DDR is in the final weeks of production, not only is this a savings in production costs, but it has the added benefit of allowing manufacturers to leave their decision about which memory to produce until almost the very last minute. This, in turn, means that during the transition from SDRAM to DDR, manufacturers can react very quickly to market trends and will have much less exposure to high levels of inventory. Of course, they will have to get rid of their existing inventory first…
In looking at the chipset roadmaps this month, the DDR vs. DRDRAM battle for the desktop seems to be all but over. Intel did reiterate their devotion to DRDRAM at IDF in February, however this should come as no surprise. It would be rather stupid for them to claim DRDRAM has no future when they currently support no other memory on their platform of the future, and despite whatever other problems they may have, the people at Intel are not stupid. It has become apparent, however, that only Intel is sticking with DRDRAM, and with P4 sales reportedly less than stellar, it is difficult to figure out where the parts that Samsung is pumping out will be sold – but possibly sales are much better than what I’ve been led to believe.
The really interesting question at this time is whether Rambus will be able to reap any financial benefits from SDRAM and DDR. With the trial against Infineon in full swing, things are not looking quite as rosy for Rambus as they did several weeks ago. According to several well publicized reports, the judge in the case has issued a preliminary judgement to limit the scope of the patent claims, and has reduced the number of allegations against Infineon from 57 to 8. The suspense will continue for another few weeks, however, as the judge has delayed the trial until April 10 to allow Rambus to collect evidence that they allege has been withheld by Infineon.
The final decision in this case could be far-reaching, and lend more weight to the decision mentioned in the article referred to last month with regards to limiting the scope of patents. After years of ever broadening definitions of what is covered by nebulous patents, the pendulum appears to be swinging the other way. Patents are intended to encourage competition by protecting inventors, but in the past few years patent law has appeared to stifle some innovation as patent owners scramble to claim ownership of anything even remotely resembling their inventions. Ironically, the very company that some have pointed to as the epitome of what is wrong with current patent laws may be involved in the pivotal case that shapes the future of IP in the U.S.
Be the first to discuss this article!